Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Introduction of Distributed Key-Value Storage “ okuyama” Kobe Digital Labo, Inc.              http://www.kdl.co.jp/
BigTable Dynamo Tokyo Ty r ant kumofs okuyama What is okuyama? Distributed Key-Value Storage
<ul><li>Distributed KVS, implemented in Java </li></ul><ul><li>Multiple data preservation form </li></ul><ul><li>Performan...
・ 100%Java -Communication Part, Control Part, Data Storage  ・ doesn’t depend on OS -Java Virtual Machine environment ・ Ver...
・ You can choose the way of preservation  to data node Main Data Node 1. Preserve all data to memory - Non-perpetuity type...
・  The scale out is possible without the system hung in both mastering nodes and the data nodes. ・  All the data shifts at...
・ Data Flow Main Master Node Main Data Node Slave Data Node Main Data Node Slave Data Node Main Data Node Slave Data Node ...
5. Function of managing collectively
・  It is not only a relation of Key-Value! >You can add Tags set (Key=“okuyama”, Tag={“oss”, ”kvs”}, Value=“Ditributed KVS...
Upcoming SlideShare
Loading in …5
×

Okuyama Summary

442 views

Published on

  • Be the first to comment

  • Be the first to like this

Okuyama Summary

  1. 1. Introduction of Distributed Key-Value Storage “ okuyama” Kobe Digital Labo, Inc.              http://www.kdl.co.jp/
  2. 2. BigTable Dynamo Tokyo Ty r ant kumofs okuyama What is okuyama? Distributed Key-Value Storage
  3. 3. <ul><li>Distributed KVS, implemented in Java </li></ul><ul><li>Multiple data preservation form </li></ul><ul><li>Performance gain by scale out </li></ul><ul><li>Composition where SPOF doesn't exist </li></ul><ul><li>Function of managing collectively </li></ul><ul><li>Unique function </li></ul>What is okuyama? -Features
  4. 4. ・ 100%Java -Communication Part, Control Part, Data Storage ・ doesn’t depend on OS -Java Virtual Machine environment ・ Verified on WindowsXP / CentOS5 series -Developed and verified on Windows -Load tested on CentOS <ul><li>Distributed KVS, implemented in Java </li></ul>
  5. 5. ・ You can choose the way of preservation to data node Main Data Node 1. Preserve all data to memory - Non-perpetuity type 2. Only preserve the data operation record to files - Perpetuity type 3. Preserve data themselves to files - Perpetuity type 2. Multiple data preservation form
  6. 6. ・ The scale out is possible without the system hung in both mastering nodes and the data nodes. ・ All the data shifts at the scale out etc. are done by the automatic operation. Main Data Node Slave Data Node Main Data Node Slave Data Node Main Data Node Slave Data Node Main Data Node Slave Data Node Main Data Node Slave Data Node Main Data Node Slave Data Node Add Node Data Shift Add 3. Performance gain by scale out
  7. 7. ・ Data Flow Main Master Node Main Data Node Slave Data Node Main Data Node Slave Data Node Main Data Node Slave Data Node Main Data Node Slave Data Node Client Slave Master Node ① Input Data 4. Composition where SPOF doesn't exist
  8. 8. 5. Function of managing collectively
  9. 9. ・ It is not only a relation of Key-Value! >You can add Tags set (Key=“okuyama”, Tag={“oss”, ”kvs”}, Value=“Ditributed KVS”); set (Key=“httpd”, Tag={“oss”, ”webserver”}, Value=“Typical WebSV”); getTagKeys(“oss”);   >Result {“okuyama”, ”httpd”} You can get all keys, resistered in same tag Data can be grouped! 6.Unique function

×